Cephalometric measures correlate with polysomnography parameters in individuals with midface deficiency

To determine the association between cephalometric measurements and polysomnographic parameters in Brazilian patients with midface deficiency. This was a primary, clinical, observational, longitudinal, retrospective, analytical, and single-center study. Forty-eight patients with midface deficiency were divided into two groups as follows: those who underwent surgically assisted rapid palatal expansion (SARME) and those who received maxillary advancement (MA). Pre- and post-operative cephalometric and polysomnography measurements were obtained. Pearson's correlation was used to verify the presence of any significant associations between PSG scores and cephalometric measurements. Associations between BMI (Body Mass Index) and AHI (Apnea Hypopnea Index) as well as arousals were observed. In the SARME group, associations between AHI and SNA, UAS and MP-H, arousals and SNA, and Co-A and MP-H were noted. Associations between AHI and Co-A, PoOr-A and MP-H, arousals and UAS, and between minimum saturation of O2 and SNA, SNB, and Co-A were observed in the MA group. This study demonstrates the alterations in the middle third of the face that were related to sleep disturbance. In addition, it shows the associations between the polysomnographic parameters and the cephalometric representations corresponding to the analyzed deformities and transverse or anteroposterior maxillary deficiencies.

Subsequently, the following angular and linear measurements were taken: SNA, the angle formed between points S-N-A; SNB, the angle formed between points S-N-B; effective jaw length, measured between Co-A points; PoOr-A, linear measurement between the line joining Po-Or points and its orthogonal plotted to point A; NPerp-A, distance from point A to a line that passes at point N and perpendicular to the PoOr plane; Upper airspace (UAS), measurement drawn from the posterior wall of the soft palate to the region of the smallest crosssection of the posterior pharyngeal wall; Lower airspace (IAS), measured from the base of the tongue (T1) and the region of the greatest narrowing with the posterior pharyngeal wall; and Hyoid bone height (MP-H), orthogonal distance from the mandibular plane (Go-Me) and the H point (Fig. 1).
Overnight type 1 polysomnography was performed in all patients before the surgery (T1) and postoperatively (T2) at Sleep Institute Clinic, Sao Paulo -Brazil. The patients were accommodated in a dark and quiet room with a comfortable bed. Electroencephalogram (EEG), submental and tibial electromyography, right and left electro-oculogram, electrocardiography, airflow (oral and nasal), respiratory effort (chest and abdomen movements), and oxyhemoglobin saturation (SatO2%) were recorded with finger pulse oximetry. A microphone was used to detect snoring and lying positions. Apnea is considered as complete cessation of breathing for ≥ 10 secs and hypopnea is considered as an abnormal respiratory event for ≥ 10 s with ≥ 30% reduction in airflow with a fall of ≥ 4% of SPO2. AHI consisted of the total number of apnea or hypopneas per hour of sleep.
Statistical analysis. All reference points were plotted at three different times with 15 days difference by the same examiner to evaluate intraobserver agreement through the intraclass correlation coefficient (ICC). The average of the values of three measurements was used for each measure. The null hypothesis was that the measurements were not concordant. Pearson's correlation test was performed to assess the relationship between the cephalometric and polysomnographic measurements for each type of surgery. Scatter plots were used for visual analysis in the case of moderate, strong, and very strong correlations. The classification proposed by DEVORE 23 was used to estimate the power of the correlation as follows: 0-0.19, very poor correlation; 0.20-0.39, weak correlation; 0.40-0.69, moderate correlation; 0.70-0.89, strong correlation; and 0.90-1.00, very strong correlation. www.nature.com/scientificreports/   www.nature.com/scientificreports/ Initially, the association between the BMI and age of the entire sample (n = 48) with polysomnographic parameters was analyzed. Then, the analysis of the associations in each group was performed. For significant correlations, the mediation effect in multiple linear regression was tested, analyzing the interaction of BMI and age in the association.
Research involving human participants and/or animals. The study was submitted and approved by the Research Ethics Committee of the Federal University of São Paulo (UNIFESP; approval number, 0178/2017). All the phases of the study strictly followed the Declarations of Helsinki for experiments involving humans.
Informed consent. Informed consent was obtained from all the participants in this study. The informed consent was previously analyzed and approved by the Research Ethics Committee of the Federal University of São Paulo (UNIFESP).

Results
The values of the cephalometric and polysomnographic measurements at T1 and T2 in patients who underwent SARME and MA are shown in supplementary online tables.

Intraclass correlation coefficient (ICC).
The ICC values were highly significant (p < 0.0001) for all the measurements, thus rejecting the null hypothesis (Table 3).
Pearson's correlation. Both age and BMI have a positive correlation with AHI, Arousals and TST SatO2 < 90% whereas they are negatively correlated with Min SatO2 (Table 4). Similar correlations were obtained when these parameters were assessed separately for SARME and MA groups (Tables 5 and 6).
The mediation regression was shown to be significant with p values < 0.05 between the association SNB/ arousals and MP-H/AHI in the SARME group when moderated by BMI, with p-values equal to 0.026 and 0.025, respectively.
For the MA group, the mediation regression showed an interaction of the BMI with the SNB/MinSatO2 association, with a p-value = 0.030 (Table 9).
There was no interaction between the AGE variable and the significant associations.

Discussion
This study was conducted to explore the association between the cephalometric measures and polysomnographic parameters in individuals with midface deficiency. Two polysomnographic parameters-pre-operative AHI and arousals were found to be associated with different cephalometric measures. Preoperative AHI was significantly    www.nature.com/scientificreports/ additional information about individuals more prone to OSAS, thus helping to rationalize the use of complementary exams. Polysomnography is considered as the gold standard for the diagnosis of OSAS, however, this is a complex, uncomfortable and expensive procedure and hence not routinely performed before orthognathic surgery for the correction of dentofacial deformities, especially in the developing countries. Furthermore, polysomnography may have the "first-night" effect (i.e., the result of the first exam), which may differ significantly from the results of subsequent exams after the patient gets adapted to it 24 . A high number of individuals with DFDs who sought surgical treatment and who, despite the presence of the symptoms of SED, had not yet been diagnosed with OSAS. This indicates that the association between DFDs and "silent" OSAS is frequent, but underdiagnosed 3 .
In the present study, patients were evaluated for the presence of DFD with surgical indications. None of the patients underwent SARME or MA for the treatment for OSAS. Basal polysomnography was performed in all patients as part of the UNIFESP craniomaxillofacial surgery outpatient care protocol. The distribution of demographic characteristics between the groups was homogeneous, with the SARME group presenting with a mean age of 30.32 years and a BMI of 26.45 kg/m 2 , while the MA group had a mean age of 31.30 years and a BMI of 26.05 kg/m 2 .
The analysis of the association between the eight cephalometric measurements and the four polysomnographic measurements was performed by Pearson's correlation coefficient. When the whole sample was analyzed, a moderate positive correlation between AHI and BMI was observed. In the study by Borges et al. 15 , no association between BMI and AHI was reported; on the other hand, other studies by Kubota et al. 25 and Tufik et al. 26 demonstrated an association between the two parameters.  Figures: a-Figure 8  www.nature.com/scientificreports/ Age is also defined as a risk factor for OSAS. However, in this study, the correlation obtained was not significant, perhaps because the overall mean age (30.79 years) was low and uniform between the two groups. Furthermore, AHI was positively associated with SNA (r = 0.499/p = 0.011) and MP-H (r = 0.573/p = 0.03) and negatively associated with UAS (r = − 0.488/p = 0.013) in the preoperative evaluation, with no significant associations with T2 measurements in the SARME groups. This finding is in agreement with the classic study by Guilleminault et al. 4 , and that by Armalaite et al. 27 . In addition, the association between AHI and MP-H has been reported by Borges et al. 15 , Silva et al. 28 and Ryu et al. 29 .
The correlation between MP-H, UAS, and AHI would translate into a more caudally located larynx, which has been associated with a risk factor for OSAS, whereas a higher UAS would translate into a larger upper airway caliber and therefore, would be expected. Lower inspiratory collapse during sleep, with lower AHI. However, a more anterior maxilla would be expected to provide larger retropalatal airspace and, consequently, a smaller AHI. This finding was not verified in the current study. Unlike the study by Parhiz et al. 30 , the average initial SNA in the present study was 82.94, and reached 81.78 postoperatively. Bretos et al. 31 observed maxillary displacement in the group that underwent SARME with Haas but not in the group that underwent SARME with Hyrax. This correlation, although significant, is not as evident in the scatter plot, and should be considered with caveats.
Arousals index was positively correlated with SNA (r = 0.410/p = 0.042), Co-A (r = 0.491/p = 0.013), and MPH (r = 0507/p = 0.010), preoperatively. Similarly, as with AHI, the translation of MP-H into the lower larynx predisposes to obstructive events with increased AHI, and consequently, to arousals. A positive correlation between SNA would not be expected, translating a more anterior maxilla and a greater number of arousals. Another unexpected positive correlation was between Co-A, effective maxillary length, and a higher rate of arousals.
The validity of such correlations may be questioned because the variation in the average Co-A was insignificant between T1 and T2 (82.70-82.12). Nevertheless, the scatter plots in this study did not suggest a significant correlation. Such associations may have occurred in isolation in this sample, but the fact that the anteroposterior dimension of the maxilla may be correlated with obstructive events and arousals in patients with transverse maxillary deficiency warrants further evaluations. One might speculate that this enlarged anterioposterior dimension could result in a more constricted maxilla thus, justifying the obstructive events and arousals.
Changes in the lateral wall of the oropharynx and the narrowing of its caliber have already been suggested by Cahali et al. 32 as a possible predisposing factor for OSAS, having served as the basis for the development of the lateral pharyngoplasty technique. Such changes in transverse length could also be associated with the development of obstructive sleep events in more cranially located anatomical sites, such as nasal cavities and nasopharynx.
Significant correlations were observed preoperatively in patients belonging to the MA group. AHI was positively correlated with CoA (r = 0.424/p = 0.044), PoOr-A (r = 0.551/p = 0.006), and MPH (r = 0.479/p = 0.021). In this case, an increase in the distance from point A to the Frankfurt line (an increase in the vertical dimension of the maxilla) was observed, which may translate into dolichofacial individuals and might be a predisposing factor for obstructive events, as suggested by Kubota et al. 25 . As mentioned previously, an increased MP-H, representing a lower-located larynx, was associated with obstructive events in patients with decreased anteroposterior maxillary dimensions. However, Co-A, which represents the effective maxillary length, was positively correlated with AHI (mean value, 83.46-88.93). Such a correlation was unexpected because its increase following surgery should have been positively correlated with the postoperative values since there was a slight increase in AHI between T1 and T2, but the persistence of this correlation was not observed. The trend line in the scatter plot is influenced by an outlier and that the other group together without such a clear correlation.
There was a negative correlation between the arousals index and the UAS (r = − 0.488/p = 0.013) ; therefore, the higher the index the lower the number of arousals per hour. Similar to the SARME group, a larger UAS provides a clearer airway without the propensity for collapse. Such an association might be expected with AHI, a fact that was not verified in this study.
The mean maxillary advancement was 7.02 mm, which was higher than that presented by Hellak et al. 33 . Furthermore, the average increase in UAS was 27.05% in the MA group; this is lower than that (37.70%) reported by Hernández-Alfaro et al. 34 . Likewise, Rosario et al. 35 observed a significant increase in UAS volume, but question the sustainability of this increase in the long term.
Despite the use of modern technological imaging resources, the delimitation of anatomical points can prove challenging and may lead to inaccuracies in the method. Some anatomical landmarks such as Porion, Condylum, and Gonius have low reliability in the inter-observer analysis, regardless of the degree of experience of the examiner 36 . Although cephalometric measurements have good reproducibility, measurement errors may occur due to difficulties in identifying certain anatomical landmarks. The values of ICC for all the cephalometric variables in this study are above 0.96 and any value above 0.9 is considered to have excellent reliability 37 .
In the present study, the Sella (S), Nasion (N), Supramentonian (B), Hyoid (H), and Mento (Me) were relatively easy to identify. However, the other demarcated points may present with relative difficulties due to the presence of overlapping images and the rotation of the patient's head. Nonetheless, despite the limitations concerning the identification of the anatomical landmarks and the potential risk of variations in the interobserver measurements, cephalometry is a highly relevant method used in medical and dental practice because it is easily acquired, has low exposure to ionizing radiation, and is cost-effective. Furthermore, this method has gained importance in detecting the skeletal changes related to various conditions, such as OSAS, and its use has extended to the detection of areas of upper airway narrowing.
Finally, negative correlations were observed between SNA, SNB, Co-A, and the minimum O2 saturation; however, they were not evident in the scatter plots. The significant correlations obtained were of moderate degree; however, MP-H may prove to be a promising cephalometric measure that can be used to demonstrate an association with OSAS, as demonstrated by Hoekema et al. 38 . Further studies that associate such measures with validated questionnaires for OSAS or excessive daytime sleepiness (Berlin or Epworth, respectively) can be sought as a way to increase the predictive value for OSAS. Moreover, studies that, through multiple regression, www.nature.com/scientificreports/ seek mathematical equations that determine the predictive value of surgeries or the predictive value for OSAS detection, as performed by Kim et al. 39 could provide relevant information.